LLMs can help with those early reads and save some time and get you into the deep reading sooner with more context. If during the deep reading you would normally jump back to a previous section to check something, it's likely faster and easier to just have a conversation with the paper (enabled with an LLM). The same would be true for that final read where you're doing logical checks.
If you use an LLM to give you a summary and leave it at that, you'll have done the equivalent of the first pass through a paper. That could be enough for you to know you want to filter it out and not do a deep reading, but you'll lose the benefits of the deeper reading. It seems like there are clear benefits and areas where LLMs can help improve that current paper reading process but if you skip (instead of just replacing with a more efficient LLM alternative) major parts of that process you'll get less out of it than you would without skipping those steps.
But having an extensive summary or table of contents generated for you to begin your review? Priceless and would have saved me so much time especially on the junk papers. There was a demo recently at work where they built a pipeline to do literature reviews (topic was not scientific, more data analysis) and generate a report. It was genuinely incredible.
I get where this is coming from, but as someone who recently did an extensive systematic literature review: you benefit from doing the work, not from getting an automatic summary. It's the little details you keep stumbling upon, that make you think "Wait a second!", that are really important. You miss them the first 100 times you come across them, but by the 101st time, you have learned something.